The Use of Multidimensional Item Response Theory Estimations in Controlling Differential Item Functioning

Differential Item Functioning (DIF), which indicates differences in item performance between individuals of equivalent ability in different groups, is the focus of this study. The aim was to investigate the effectiveness of Multidimensional Item Response Theory (MIRT) ability estimation in controlling DIF under different simulation conditions. For the purpose of the study, different experimental conditions were created by varying the number of dimensions, DIF percentage, DIF size and DIF type. Depending on the changed conditions, Item Response Theory and MIRT estimations were performed; the significance of the differences in Type I error and power ratios obtained from the DIF analysis of the data sets obtained from the estimations were tested with Factorial ANOVA. This study shows that in tests with DIF, taking into account the ability level related to the structure by estimating with MIRT provides a correction in the ability estimation and controls the DIF effect. In the Uniform DIF type, Type 1 error increases as the DIF size and percentage increase, and multidimensional estimation provides more correction on power ratios as the DIF size and number of dimensions increase. Whereas in the non-uniform DIF type, Type 1 error increases as DIF size increases; multidimensional estimation provides more correction on power ratios as DIF size, DIF percentage and number of dimensions increase. In this context, it is recommended that practitioners should perform MIRT estimation in the presence of DIF and conduct DIF investigations based on the corrected ability estimation taking into account the ability scores related to the structure.